RFR (S/M): 8136678: Implement adaptive sizing algorithm for IHOP

Thomas Schatzl thomas.schatzl at oracle.com
Thu Nov 19 10:56:17 UTC 2015


Hi Jon,

  thanks for your review.

On Tue, 2015-11-17 at 11:35 -0800, Jon Masamitsu wrote:
> Thanks for the changes.
> 
> http://cr.openjdk.java.net/~tschatzl/8136678/webrev.0_to_1/src/share/vm/gc/g1/g1IHOPControl.cpp.frames.html
> 
>  185 static double percentage_of(double quantity, double base_quantity) {
> 
> looks a lot like
> 
> share/vm/gc/g1/g1RemSetSummary.cpp
> 
> static double percent_of(size_t numerator, size_t denominator) {
> 
> Time to add a function to globalDefinitions?

I knew I had written that before. Fixed.

> Minor issues.
> 
> http://cr.openjdk.java.net/~tschatzl/8136678/webrev.0_to_1/src/share/vm/gc/g1/g1IHOPControl.hpp.frames.html
> 
> You can ignore this but
> 
>  103 // depending on predictions of current allocation rate and time periods between
> 
> might be better as
> 
>  103 // based on predictions of current allocation rate and time periods between
> 
> Something got jumbled up?
>  117   // non-young gen occupancy compared against at the end of GC, but we need to
> 

Fixed. Thanks.

This change incorporates some comments from Erik too, in particular the
observation that G1AdaptiveIHOPControl::calculate() is not required at
all (and the cached _current_threshold value) and can be inlined in
G1AdaptiveIHOPControl::get_conc_mark_start_threshold() as well.

New webrevs:

http://cr.openjdk.java.net/~tschatzl/8136678/webrev.1_to_2/ (diff)
http://cr.openjdk.java.net/~tschatzl/8136678/webrev.2/ (full)

Thanks,
  Thomas




More information about the hotspot-gc-dev mailing list